Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

European Congress on Computational Methods in Applied Sciences and Engineering

ECCOMAS 2000
Barcelona, 11-14 September 2000
c
ECCOMAS

FINITE DIFFERENCE HARTREE-FOCK METHOD FOR


DIATOMIC MOLECULES
J. Kobus
Instytut Fizyki, Uniwersytet Mikolaja Kopernika, ul. Grudziadzka 5/7, 87-100 Toru
n,
Poland e-mail: jkob@phys.uni.torun.pl

Key words: Diatomic molecules, Finite-derence Hartree-Fock method, Basis set error,
Universal basis sets, Electric multipole moments
Abstract. This paper presents the nite dierence Hartree-Fock method for diatomic
molecules together with several recent examples of its usage. This approach has been
applied to assess a basis set truncation and superposition errors, to develop sequences of
universal even-tempered Gaussian basis sets supporting Hartree level of accuracy and to
estimate parity non-conservation eects in the Y bF system.

J. Kobus

INTRODUCTION

Nearly a half century ago C. Roothaan published his renowned paper which paved the
way for a routine application of quantum mechanics to molecular systems [1]. Roothaan
proposed to expand a wave function in a given set of basis functions in order to transform
the Schrodinger equation into its algebraic equivalent. In particular, molecular orbitals of
the Hartree-Fock method can be written as linear combinations of atom-centred basis set
functions which makes the approach applicable to molecular systems of any symmetry.
There is, however, a price to be paid for this generality: basis sets used are inevitably
nite and thus far from being complete and their choice (providing an adequate level of
accuracy) has turned out to be dicult. As a consequence all calculations performed
within the algebraic approach are ridden by basis set truncation. Ever since computational quantum chemists have been preoccupied with the task of basis sets selection and
development (see, for example, [2]).
Over the last four decades one has witnessed the extensive usage of the algebraic
Hartree-Fock method accompanied by the continuous eorts to develop basis sets appropriate for the description of various molecules and their properties. Diculties in
controlling the basis set truncation error and dependence of the calculated properties on
the basis sets quality mean that in principle ab initio calculations performed within the
algebraic Hartree-Fock approach assume in fact a quasi-empirical character where basis
set truncation errors are treated as a part of the model employed [3].
Numerical methods have always been prevalent in solving the Hartree-Fock equations
of atomic electronic structure problems. It has been just the opposite in the case of
their molecular counterparts where the lack of spherical symmetry rendered the methods
computationally untractable or dicult to implement. Diatomic molecules are the only
molecular systems which have routinely been treated numerically. This is due to the fact
that for such systems the computational complexity of the problem can be reduced and
an accurate description of atomic centres is possible by a suitable choice of the coordinate
system.
Since the mid 1970s there have been several attempts to devise basis-set-independent
numerical approaches [4] (and references therein). The rst of these was due to McCullough who proposed to solve the (multi-conguration) Hartree-Fock equations using the
partial-wave expansion method [5]. At the early 1980s Laaksonen et al. developed the
nite dierence numerical Hartree-Fock method [6] (rened later by Kobus et al. [7])
and at the end of the decade Heinemann et al. [8, 9] implemented the nite element
Hartree-Fock method.
In recent years the nite dierence Hartree-Fock method has been used to provide
Hartree-Fock-limit values of total energies and multipole moments for a wide range of
closed and open-shell diatomic molecules: H2 [10, 11], He2 , LiH, Be2 , BH [10, 12], HF
[12, 13], BeF , BO, CN, N2+ [14], N2 , CO, BF , NO + , CN [15], CS [16], AlF , GaF
[17], InF , T lF [18], MgF , CaF , SrF [19], HCl, HBr, HI [13].

J. Kobus

At present the main and probably the most important application of the nite dierence Hartree-Fock method is in the developing the universal sequences of even-tempered
Gaussian basis sets and in assessing their quality [20, 21]. It has also been used to study
the dependence of the basis set truncation errors on the number and type of basis set
functions in order to test basis set extrapolation techniques[11, 22].
This paper aims at presenting a brief overview of the nite dierence Hartree-Fock
method together with some examples of its application. The introduction is followed
by Section 2 in which the essential features of the method and its implementation are
described. Section 3 presents a number of recent applications of the method. In particular, the method has been applied (i) to estimate the accuracy of spectroscopic constants
calculated for several hydrogen halide molecules with relativistic and correlation eects
included, (ii) to analyze a dependence of basis set truncation errors on internuclear distance and the usefulness of a complete-basis-set technique for the beryllium dimer and (iii)
to calibrate sequences of even-tempered Gaussian basis functions that support Hartree
level of accuracy. The nite dierence Hartree-Fock solutions for the Y bF have been used
to calculate the strength of parity non-conservation eects in this system.
2

METHOD AND ALGORITHM

A detailed description of the nite dierence Hartree-Fock method can be found in [6]
and [7]. In this section a brief description of the method and its implementation is given.
The restricted open-shell Hartree-Fock method (cf. Hurley[23]) leads to the following
equations for molecular orbitals and potentials:

1
2 a = (Vn + VC Vxa Ea )a +
Eab b
2
b=a

(1)

with
Vn =

Z1 Z2
,
r1
r2

VC =

VCa ,

Vxa =

Vxab ,

(2)

b=a

where
2 VCa = 4a a ,

2 Vxab = 4a b .

(3)

For diatomic molecules the choice of the prolate spheroidal coordinate system
= (r1 + r2 )/R
= (r1 r2 )/R
azimuth angle

1 <
1 1
0 2

(4)

J. Kobus

allows one to express orbitals and potential functions in the form f (, )eim and to reduce
the complexity of the Hartree-Fock equations by treating the variable analytically. In
the above equations ri , (i = 1, 2), denotes the radial distance from the nucleus Zi and
R the internuclear distance. To allow for a more accurate description of orbitals and
potentials in the vicinity of the nuclei, the prolate spheroidal coordinates are transformed
into the (, , ) variables where
= cosh1
= cos1

0
0 .

(5)

It can readily be shown that the Hartree-Fock eqs (1) and (3) can be rewritten in the
form


2

2
+ C() 2 + D()
+ E(, ) u(, ) = F(, ).
(6)
A() 2 + B()

This is a second-order partial dierential equation of the elliptical type in two variables and
its solutions are sought on a rectanglar region [0, ] [0, ], where cosh( ) = 2r /R
with a suitable value for the practical innity, r . This value must be chosen large
enough to guarantee that the boundary conditions derived from the asymptotic form of
the molecular orbitals and the potentials can be applied.
The partial dierential equations are discretized by the eighth-order central dierence
stencil. The resultant large and sparse system of linear equations for the values of orbitals and potentials at grid points are then solved by the successive overrelaxation method
(SOR) [6, 24] or its multicolour variant (MCSOR) [25] which is well suited for computers
with vector processors. To solve the Hartree-Fock equations the self-consistent iterative
scheme is used and thus both the self-consistent and successive overrelaxations iterations
are interwoven. By default in any self-consistent step every orbital and potential function
undergoes 10 (MC)SOR iterations and depending on the case from several hundered
to several thousend scf iterations are required to solve the equations to 10-12 digits accuracy. Since about 90-95% of all the computing time is consumed by solving these linear
equations the eciency of the approach depends solely on the ecacy of their solver.
That is why a great deal of eort was devoted to a careful coding of these algorithms.
Their performance on a range of computers is illustrated in Table 1.
3

APPLICATIONS

In many instances the Hartree-Fock results are not of the special importance by themselves since the model is usually only the rst step in a series of more rened calculations
including relativistic and correlation eects. Nevertheless these results can provide a valuable reference and be used to estimate the accuracy of nite basis set calculations in cases
where the algebraic Hartree-Fock results can be compared with the numerical ones.
Recently Styszy
nski has used the MOLFDIR package of programmes [26] to examine
the inuence of relativistic core-valence correlation eects on molecular properties of a
4

J. Kobus

number of hydrogen halide molecules [13]. In order to assess the basis set quality and
determine the dependence of the accuracy of the spectroscopic constants on the basis set
truncation error the algebraic (nonrelativistic) Hartree-Fock results were compared with
their nite dierence counterparts (Table 2). It has turned out that for all systems studied
the equilibrium bond lengths could be calculated with an accuracy of up to 5 m
A, the
1
harmonic frequences 10 cm and the dissociation energies 1 kcal/mol. These values
may be considered the minimal errors of the calculated properties due to the deciency
of the basis sets used.

SOR
MCSOR

Cray DEC IBMa SGIb


Sunc
Y-MP 8400
SP2
PCh E6000
19.1
5.3
7.3
6.2
7.7
(25.7) (93.0) (67.2) (79.1) (64.1)
3.3
12.1
8.5
9.5
15.3
(149) (40.5) (57.7) (51.6) (32.3)

PCd
5.5
(89.2)
8.9
(55.1)

IBM SP2: IBM RS6000/590, 66MHz wide node


SGI PCh: SGI Power Challenge R8000, 90MHz, 4MB cache
c
Sun Enterprise 6000, UltraSparc 167MHz, 0.5MB cache
d
Pentium III 500 MHz, Linux Red Hat 6.1, g77 compiler
b

Table 1: Performance of the (MC)SOR algorithm on a range of computers. A single processor CPU
times (in seconds) required to perform 100 (MC)SOR iterations on a 294 298 grid are given with the
corresponding MFLOPS rates in parentheses (in every (MC)SOR iteration 56 oating-point operations
were required to relax the solution at a single grid point).

R e (
A)
e (cm1 )

De (kcal/mol)

method
fdHF
fbHF

fdHF
fbHF

fdHF
fbHF

HF
HCl
HBr
HI
0.8970 1.2640 1.407 1.606
0.8991 1.2674 1.408 1.611
-0.0021 -0.0034 -0.001 -0.005
4473.9 3142.9 2796
2463
4465.8 3135.6 2795
2457
8.1
7.3
1
6
101.7
82.2
69.8
59.0
101.4
81.2
69.3
58.0
0.3
1.0
0.5
1.0

Table 2: Equilibrium bond lengths (Re ), harmonic frequencies (e ) and dissociation energies (De ) for
the algebraic (f bHF ) and nite dierence (f dHF ) Hartree-Fock calculations. denotes the dierence
between the latter and the former values.

J. Kobus

Petersson and Shirley [27] have used the Hartree-Fock interaction energy to construct
the potential energy curve for the beryllium dimer with the correlation corrections included. The corrections were calculated within the Mller-Plesset perturbation expansion
in the framework of the molecular natural pair orbitals. The contributions to the HartreeFock and correlation energy were extrapolated to the limit of a complete basis set. As
a result the remaining discrepances between the theoretical and experimental potental
energy curve were brought down to about 100Hartree (Fig. 1). In this case the nite difference Hartree-Fock method could be used to check the accuracy of the potential energy
curve at the Hartree-Fock level. As can be seen from the gure the dierence between the
algebraic and nite dierence calculations is not at all constant, especially in the vicinity
of the equlibrium bond distance. This dierence can be attributed to the basis set superposition error and the deciencies of the complete-basis-set extrapolation technique.
If the correct values were used for the Hartree-Fock interaction energy the discrepancy
between the theoretical potential energy curve and experimentl data would change considerably. The remaining discrepancies could be used to examine the convergence pattern
of the Mller-Plesset perturbation expansion and the application of the complete-basis-set
approach to estimate correlation corrections. Since the overall accuracy of the theoretical
description of the potential energy curve is quite high it looks as if the inadequacy of the
description at the Hartree-Fock level were at least to some extent covered up by the
correlation corrections.
The nite dierence Hartree-Fock method provides high precision solutions of the
Hartree-Fock equations for diatomic molecules with which the results of matrix HartreeFock calculations can be compared and the basis set truncation error estimated. It was
demonstrated that molecular basis sets consisting of sequences of even-tempered Gaussian
functions can be systematically constructed so as to support total Hartree-Fock energies of
an accuracy approaching the Hartree level [20, 21], [19] (and references therein). In the
course of constructing these basis sets the convergence of the total energy with respect to
(i) the subset of functions associated with a given symmetry on a given expansion centre,
(ii) the number of symmetries included on a given expansion centre and (iii) the number
of expansion centres were examined. It was found desirable to include bond-centre basis
functions in order to speedup the convergence. Thus obtained basis sets have been recently shown to be able to support the accurate calculation of multipole moments [19].
Total energies and electric dipole, quadrupole, octopole and hexadecapole moments of
some closed- and open-shell uorides containing increasingly heavy atom are displayed in
Tables 3 and 4.
To date the T lF molecule is the largest diatomic system treated by both the algebraic
and nite dierence Hartree-Fock methods. The details of a tedious basis set construction
process can be found in [18, 30] and its summary is graphically illustrated in Fig. 2. This
system attracted a lot of interest recently because it had been advocated as the ideal
molecular system to look for parity non-conservation and time-reversal violation eects
[31, 32]. Due to the weak interactions these eects are present (and in principle could be
6

J. Kobus
0.0005

energy differences (Hartree)

0.0004

0.0003

0.0002

0.0001

-0.0001
4

10

11

12

internuclear distance (bohr)

Figure 1: Basis set truncation and superposition errors of the beryllium dimer potential energy curve. >
marks the dierence in the Hartree-Fock interaction energy between the nite basis set calculations (with
complete-basis-set and basis set superposition errors included) and the nite-dierence results [10]. The
line with  points shows the dierence between the experimental potential energy curve and the results
of the complete-basis-set full CI calculations[28] and the line with  shows the full CI results corrected
for the basis set truncation error at the Hartree-Fock level. The experimental bond distance is 4.7 bohr.
0.1

0.01

0.001

0.0001

1e-05
200

400

600

800
1000
1200
number of basis functions

1400

1600

1800

Figure 2: Development of a sequence of uncontracted even-tempered Gaussian basis sets for T lF . Truncation basis set errors (in Hartree) calculated as the dierence between the algebraic and nite dierence
Hartree-Fock total energy plotted as a function of the basis set size [18, 30].

J. Kobus

Method

EHF
fdHF
0.8713 3.9523 4.5293 3.6677 124.168 779 2
fbHF
0.8713 3.9522 4.5290 3.6732 124.168 777 0

0.0000 0.0001 0.0003


0.0055
2.2
AlF b
fdHF 1.3227 5.8863 5.9769 7.2501 341.488 382 8
fbHF 1.3228 5.8859 5.9754 7.2482 341.488 369 5

0.0001 0.0004 0.0015 0.0019


13.3
GaF c
fdHF 2.2235 5.9097 6.1715 9.7471 2022.836 882 2
fbHF 2.2255 5.9048 6.1630 9.7719 2022.836 862 3

0.0020 0.0049 0.0085


0.0248
19.9
d
InF
fdHF 2.8551 7.8301 8.2079 16.3040 5839.729 193 5
fbHF 2.8574 7.8218 8.1887 16.2744 5839.729 178 0

0.0023 0.0083 0.0192 0.0296


15.5
e
T lF
fdHF 3.2138 8.6314 9.0890 20.3563 19061.376 349
fbHF 3.2168 8.6202 9.0615 20.3067 19061.376 288

0.0030 0.0112 0.0275 0.0496


61
a
[169 193; 40], [B : 26s18p18d18f F : 26s18p18d18f ; bc : 23s13p13d14f ] [17]
b
[169 193; 40], [Al : 28s20p20d20f ; F : 26s18p18d18f ; bc : 22s15p17d17f ] [17]
c
[295 295; 40], [Ga : 48s24p24d24f ; F : 30s15p15d15f ; bc : 27s12p14d15f ] [17, 18]
d
[595 595; 40], [In : 56s28p28d28f 28g; F : 30s15p15d15f ; bc : 51s23p24d25f 25g] [18]
e
[595 595; 45; ], [T l : 58s29p29d29f 29g; F : 30s15p15d15f ; bc : 53s24p26d25f 25g] [18]
BF a

Table 3: Dipole (), quadrupole (), octopole () and hexadecapole () moments for some closedshell diatomic uorides calculated relative to the geometrical centre. 2n -pole moment is given in Debye

Angstromn units, EHF in Hartree and , the dierence between Ef bHF and Ef dHF , in Hartree.

J. Kobus

Method

EHF
BeF
fdHF 1.2727 4.1622 5.8728 7.5857 114.172 685 040
fbHF 1.2726 4.1624 5.8727 7.5843 114.172 683 8
0.0001
0.0002 0.0001 0.0014
1.2
b
MgF
fdHF 3.1005 6.7716 9.3090 17.1685 299.159 119 304
fbHF 3.1005 6.7715 9.3074 17.1509 299.159 112 9

0.0000 0.0001 0.0016 0.0176


6.4
c
CaF
fdHF 2.6450 10.6872 14.2460 26.7662 776.329 958 335
fbHF 2.6451 10.6871 14.2432 26.7324 776.329 956 0

0.0001 0.0001 0.0028 0.0338


2.3
SrF d
fdHF 2.5759 11.8026 14.1228 26.1455 3231.119 630 232
fbHF 2.5760 11.8024 14.1225 26.1506 3231.119 629 2

0.0001 0.0002 0.0003


0.0051
1.1
a
[319 415; 40], [Be : 32s17p17d17f ; F : 30s15p15d15f ; bc : 9s8p9d8f ] [29]
b
[319 415; 40], [Mg : 32s17p17d17f ; F 30s15p15d15f ; bc : 9s9p10d8f ] [29]
c
[349 499; 80], [Ca : 48s27p27d27f ; F : 30s15p15d15f ; bc : 10s9p9d11f ] [29]
d
[391 547; 80], [Sr : 56s31p31d31f 31g; F : 30s15p15d15f ; bc : 9s9p9d9g] [29]
a

Table 4: Multipole moments for some open-shell diatomic uorides calculated relative to the geometrical
centre. See Table 3 for denitions of symbols and units.

observed) in every process in atomic, molecular and solid state physics. However, because
of their smallness, these interactions can only be observed in heavy atoms and molecules
where high electron density, large nucleus volume and huge elds enhance the eects by
orders of magnitude [33, 34] (and references therein).
It can be shown that the presence of the parity non-conservation and time-reversal
violation eects in molecules leads to the eective interaction of the form [31, 34]

He = dN
where d is a coupling constant characterizing the strength of the interaction, N is the

nuclear spin operator and a unit vector dening the direction of the internuclear axis.
For the volume eect the coupling constant is
dV = (2/3)dp|(0)| R = dp XR
where dp is the protons intrinsic electric dipole moment, (0) is the gradient of the
electron density at the Tl nucleus and R is a nuclear quantity dened by its structure.
The subscript indicates that only the component in the direction of the internuclear
axis has a nonvanishing expectation value. The molecular orbitals of the T lF obtained
by means of the nite dierence (nonrelativistic) Hartree-Fock approach were used to
9

J. Kobus

calculate the X factor. This fully numerical value is compared with algebraic calculations
in Table 5. It is twice as large as the very recent value of Parpia [35] and compares
very well with the nonrelativistic value of Laerdahl [36, 37] which was estimated from the
relativistic results using a relativistic enhancement ratio of Kozlov and Labzovsky [38].
The good agreement in the latter case may be attributed as Laerdhal et. al point out in
their paper to the fact that the result was obtained using basis sets optimized to make
the Hellmann-Feynman theorem satised and thus making the forces on the T l nucleus
vanish. The diculty of calculating the X factor can be appreciated when the individual
orbital contributions are examined (Table 6). The inner orbitals contribute very little to
the X value as result of a nearly perfect cancellation of right- and left-hand-side orbital
density derivatives in a sharp contrast to the ve highest orbitals which contribution
amount to just over 93% of the value.
Xrel
Xnrel
Xrel /Xnrel
Coveney, Sandars (1983)[32] 2091
294.5
7.1
Parpia (1997)[35]
7319
648.8
11.3
Laerdahl et al. (1997)[36, 37] 8747
1130.0
7.7
Moncrie et al. (1998)[18]
5920
1138.5
5.2
8083
1138.5
7.1

the relativistic enhancement ratio taken from [38]

the relativistic enhancement ratio taken from [32]


Table 5: Relativistic and nonrelativistic values of the total X factor (in au) for the

205

Tl

19

F molecule.

CONCLUSIONS

The paper presented the nite dierence Hartree-Fock method and its usage. It was
demonstrated that the method can routinely be used to provide Hartree-Fock-limit values
of total energies and multipole moments for a wide range of diatomic molecules. These
results can in turn be employed to assess basis set truncation and superposition errors
present in algebraic approaches, to calibrate basis sets and to devise and test various
extrapolation techniques.
ACKNOWLEDGEMENT
This work has been, in part, sponsored by KBN grant 8 T11F 028 18. Support towards
travel expenses from Uniwersytet Mikolaja Kopernika is gratefully acknowledged.

10

J. Kobus

orbital
(0)+
(0)
Xf d [18] Xf b [35]
17
3 312.82
4 839.01 1 598.24
856.80
16
1 134.37
577.92 582.71 296.80
15
2.92
2.50
0.44
0.40
14
56.61
61.44 52.85 31.59
13
125.05
127.23
264.16
193.06
12
64 131.63
63 976.69 162.26 114.76
11
0.01
0.00
0.01
0.00
10
0.03
0.03
0.00
0.00
9
13.00
13.00
27.23
29.63
8
0.00
0.00
0.00
0.00
7
339 679.53
339 682.37
2.98 14.24
6
0.00
0.00
0.00
0.00
5
12.15
12.15 25.45
1.35
4
1 357 846.37 1 357 886.23
41.75
11.64
3
51.54
51.51
107.92
84.71
2
6 008 664.09 6 008 570.93 95.56 78.89
1
54 057 348.50 54 057 363.26
15.46
10.42
Table 6: Orbital contributions to the X factor for the 205 Tl19 F molecule. (0) denotes the right- and
left-hand-side orbital density gradient along the internuclear axis at the T l location calculated with the
nite dierence (Xfd ) and nite basis set (Xfb ) Hartree-Fock method.

11

J. Kobus

REFERENCES
[1] C. C. J. Roothaan, New developments in molecular orbital theory, Rev. Mod. Phys.
23, 6989 (1951).
[2] P. R. Taylor, Accurate Calculations and Calibration, In , B. O. Roos, ed., Lecture
Notes in Quantum Chemistry 58, 325412 (Springer-Verlag, Berlin, 1992).
[3] S. Huzinaga, Basis sets for molecular calculation, Comput. Phys. Rep. 2, 279339
(1985).
[4] P. Pyykko, Fully numerical calculations for diatomic systems, In Numerical Determination of the Electronic Structure of Atoms, Diatomic and Polyatomic Molecules,
M. Defranceschi and J. Delhalle, eds., NATO ASI Series, Series C: Mathematical and
Physical Sciences 271, 161175 (Kluwer Academic Publishers, Dordrecht, 1989).
[5] E. A. McCullough Jr., Numerical Hartree-Fock methods for diatomic molecules: a
partial-wave expansion approach, Comput. Phys. Rep. 4, 265312 (1986).
[6] L. Laaksonen, P. Pyykko, and D. Sundholm, Fully numerical Hartree-Fock methods
for molecules, Comput. Phys. Rep. 4, 313344 (1986).
[7] J. Kobus, L. Laaksonen, and D. Sundholm, A numerical Hartree-Fock program for
diatomic molecules, Comput. Phys. Commun. 98, 346358 (1996).
[8] D. Heinemann, B. Fricke, and D. Kolb, Solution of the Hartree-Fock-Slater equations
for diatomic molecules by the nite element method, Phys. Rev. A 38, 49945001
(1988).
[9] D. Heinemann, A. Rosen, and B. Fricke, Solution of the Hartree-Fock equations for
atoms and diatomic molecules with the nite element method, Physica Scripta 42,
692696 (1990).
[10] J. Kobus, Finite-dierence versus nite-element methods, Chem. Phys. Lett. 202,
712 (1993).
[11] F. Jensen, The basis set convergence of the Hartree-Fock energy for H2 , J. Chem.
Phys. 110, 66016605 (1999).
[12] A. Halkier, W. Klopper, T. Halgaker, and P. Jorgensen, Basis-set convergence of
the molecular electronic dipole moment, J. Chem. Phys. 111, 44244430 (1999).
[13] J. Styszy
nski, Relativistic core-valence correlation eects on molecular properties of
the hydrogen halide molecules, Chem. Phys. Lett. 317, 351359 (2000).

12

J. Kobus

[14] J. Kobus, D. Moncrie, and S. Wilson, A comparison of nite basis set and nite
dierence Hartree-Fock calculations for the open-shell (X 2 + ) species BeF, BO, CN
and N+
2 , Mol. Phys. 96, 15591567 (1999).
[15] D. Moncrie, J. Kobus, and S. Wilson, A universal basis set for high precision
electronic structure studies, J. Phys. B: At. Mol. Opt. Phys. 28, 45554557 (1995).
[16] J. Kobus, D. Moncrie, and S. Wilson, A comparison of nite basis set and nite
dierence methods for the ground state of the CS molecule, J. Phys. B: At. Mol.
Opt. Phys. 27, 28672875 (1994).
[17] J. Kobus, D. Moncrie, and S. Wilson, A comparison of nite basis set and nite
dierence Hartree-Fock calculations for the BF, AlF and GaF molecules, Mol. Phys.
86, 13151330 (1995).
[18] D. Moncrie, J. Kobus, and S. Wilson, A comparison of nite basis set and nite
dierence Hartree-Fock calculations for the InF and T lF molecules, Mol. Phys.
93, 713725 (1998).
[19] J. Kobus, D. Moncrie, and S. Wilson, A comparison of the electric moments obtained from nite basis set and nite dierence Hartree-Fock calculations for diatomic
molecules, (submitted for publication).
[20] D. Moncrie and S. Wilson, Finite basis set versus nite dierence and nite element
methods, Chem. Phys. Lett. 209, 423426 (1993).
[21] D. Moncrie and S. Wilson, On the accuracy of the algebraic approximation
in molecular electronic structure calculations. III. Comparison of matrix HartreeFock and numerical Hartree-Fock calculations for the ground state of the nitrogen
molecule, J. Phys. B: At. Mol. Opt. Phys. 26, 16051616 (1993).
[22] F. Jensen, private communication.
[23] A. C. Hurley, Introduction to the Electron Theory of Small Molecules (Academic
Press, London, 1976).
[24] J. Stoer and R. Bulirsch, Introduction to numerical analysis (Springer-Verlag, New
York, 1980).
[25] J. Kobus, Vectorizable algorithm for the (multicolour) successive overrelaxation
method, Comput. Phys. Commun. 78, 247255 (1994).
[26] L. Visscher, O. Visser, P. J. C. Aerts, H. Merenga, and W. C. Nieuwpoort, Relativistic quantum chemistry: the MOLFDIR program package, Comput. Phys. Commun.
81, 120144 (1994).
13

J. Kobus

[27] G. A. Petersson and A. Shirley, The beryllium dimer potential, Chem. Phys. Lett.
160, 494500 (1989).
[28] G. A. Petersson, A. Bennett, T. G. Tensfeldt, M. A. Al-Laham, A. Shirley, and J.
Mantzaris, A complete basis set chemistry. I. The total energies of closed-shell atoms
and hydrides of the rst-row elements, J. Chem. Phys. 89, 21922218 (1988).
[29] J. Kobus, D. Moncrie, and S. Wilson, A comparison of nite basis set and nite
dierence Hartree-Fock calculations for the open-shell (X2 + ) species BeF, MgF,
CaF and SrF, Mol. Phys. 98, 401408 (2000).
[30] S. Wilson, D. Moncrie, and J. Kobus, TlF(1 + ): Some preliminary electronic structure calculations, Technical Report No. RAL-94-082, Rutherford Appleton Laboratory, Chilton, Oxon (1994) .
[31] E. A. Hinds and P. G. H. Sandars, Electric hyperne structure of TlF, Phys. Rev.
A 21, 471479 (1980).
[32] P. V. Coveney and P. G. H. Sandars, Parity- and time-violating interactions in
thallium uoride, J. Phys. B: At. Mol. Opt. Phys. 16, 37273740 (1983).
[33] A. M
artensson-Pendrill, Calculation of P- and T-violating properties in atoms and
molecules, In Methods in Computational Chemistry: Atomic and Molecular Properties, S. Wilson, ed., 5, 99156 (Plenum Press, New York, 1992).
[34] D. Cho, K. Sangster, and E. A. Hinds, Search for time-reversal-symmetry violation
in thallium uoride using a jet source, Phys. Rev. A 44, 27832799 (1991).
[35] F. A. Parpia, Electric-dipole hyperne matrix elements of the ground state of the
TlF molecule in the Dirac-Fock approximation, J. Phys. B: At. Mol. Opt. Phys. 30,
39834001 (1997).
[36] J. K. Laerdahl, T. Saue, and K. Fgri, Jr., Ab initio study of the PT-odd interactions in thallium uoride, Phys. Rev. Lett. 79, 16421645 (1997).
[37] H. Quiney, J. K. Laerdahl, K. Fgri,Jr., and T. Saue, Ab initio Dirac-Hartree-Fock
calculations of chemical properties and PT-odd eects in thallium uoride, Phys.
Rev. A 57, 920944 (1998).
[38] M. G. Kozlov and L. N. Labzovsky, Parity violation eects in diatomics, J. Phys.
B: At. Mol. Opt. Phys. 28, 19331961 (1995).

14

You might also like